A Hornblende Basalt from Western Mexico: Water-saturated Phase Relations Constrain a Pressure---Temperature Window of Eruptibility

Transcription

1 JOURNAL OF PETROLOGY VOLUME 45 NUMBER 3 PAGES 485± DOI: /petrology/egg091 A Hornblende Basalt from Western Mexico: Water-saturated Phase Relations Constrain a Pressure---Temperature Window of Eruptibility J. BARCLAY* AND I. S. E. CARMICHAEL DEPARTMENT OF EARTH AND PLANETARY SCIENCE, UNIVERSITY OF CALIFORNIA, BERKELEY, CA , USA RECEIVED AUGUST 25, 2001; ACCEPTED JULY 31, 2003 Trachybasalt scoria from a cinder cone near the Mexican volcanic front contain phenocrysts of olivine with chromite inclusions, apatite, augite and hornblende, with microphenocrysts of plagioclase. The water-saturated phase relations reproduce the phenocryst assemblage between 1040 C and 970 C with water contents of between 25 and45% ( MPa). The absence of biotite phenocrysts in the scoria places a tight constraint on the pressure---temperature conditions of phenocryst equilibration, as there is only a small zone where biotite does not accompany hornblende in the experiments. Diluting the fluid phase with CO 2 changes the composition of the olivine, indicating that CO 2 was only a minor component of the fluid of the scoria. Hornblende is stable to 1040 C at oxygen fugacities of NNO 2 (where NNO is the nickel---nickel oxide buffer), but at lower oxygen fugacities, the upper limit is 990 C. There is a progressive increase in crystallinity in experimental runs as both pressure and temperature decrease. Isobaric plots of crystallinity show that the onset of hornblende crystallization involves a reaction relation, and also results in a marked vol. % increase in crystallinity. Ascending hydrous magmas intersecting the cooler crust could be trapped there by the large increase in crystallinity accompanying the isobaric crystallization of hornblende. KEY WORDS: crystallization; eruptibility limit; experiments; hornblende trachybasalt; Mexico INTRODUCTION Water has a profound influence not only on the temperature and depth of melting of arc-related magmas (e.g. Kawamoto & Holloway, 1997; Prouteau et al., 2001) but also on melt dynamics, influencing such processes as mixing, assimilation and differentiation (e.g. Gaetani et al., 1993; Sisson & Grove 1993; Thomas & Tait, 1997; Hort, 1998), and ultimately controls the eruptive behaviour of the magma on the Earth's surface (e.g. Sparks et al., 1994; Bower & Woods, 1997). Volcanic arcs associated with subduction, especially beneath a continental plate, are the sites of the most diverse range of magmatism found on Earth, exemplified by variations in composition and eruptive style. Estimating the water content of subduction-related magmas is therefore a first-order problem in the petrogenesis of volcanic arcs. Estimates have been obtained from the composition of melt either trapped in phenocrysts within lavas or preserved in quenched glasses from submarine eruptions (Anderson, 1974, 1979; Sisson & Layne, 1993), or by using experimental phase equilibria to reproduce the phenocryst assemblage, the phenocryst composition and the phenocryst abundance of a lava (e.g. Rutherford et al., 1985; Blatter & Carmichael, 1998, 2001; Moore & Carmichael, 1998). Such studies have unequivocally demonstrated the importance of water in controlling many aspects of subduction-related magmas. Despite this, there are few examples of high-h 2 O basalts that are erupted containing primary hydrous phases (the mineralogical evidence for high water content), with extraordinarily few occurrences documented worldwide [Bogoslof, Alaska, Arculus et al. (1977); Kick 'em Jenny, Lesser *Corresponding author. Present address: School of Environmental Sciences, University of East Anglia, Norwich NR4 7TJ, UK. Fax: 44 (0) j.barclay@uea.ac.uk Journal of Petrology 45(3) # Oxford University Press 2004; all rights reserved

2 JOURNAL OF PETROLOGY VOLUME 45 NUMBER 3 MARCH 2004 Antilles, Sigurdsson & Shepherd (1974); Cerro la Pilita, Western Mexico, Luhr & Carmichael (1985); Mount Lamington, Papua New Guinea, Arculus et al. (1983)]. This is in spite of the almost ubiquitous presence of amphibole in high-h 2 O subduction-related magmas as a whole. This would suggest that either (1) high H 2 O content magmas are largely confined to more evolved magmas or (2) there is a correlation between high H 2 O content and magma eruptibility in mafic magmas. The detailed experimental investigation of the role of P---T---H 2 O on mineral stabilities and crystallinities in hydrous basaltic magmas currently provides the only way to investigate this relationship in detail. The unique information provided on phenocryst assemblage, composition and abundance provides information of relevance not only to the P---T path of the magma but also to its likely rheological evolution, a factor that can directly control the probability of eruption. Unfortunately, complete experimental studies of hydrous mafic magmas are comparatively rare. In this paper we explore the conditions (P---T ) of water-saturated phase equilibria for one of the rare hornblende-bearing basaltic magmas (the Cerro la Pilita trachybasalt) and also the effect of CO 2 ---H 2 O on the phase assemblages. By plotting the modal abundance of crystals (referred to subsequently as the crystallinity) on the phase diagram we investigate more fully the role that water played in both magma ascent and eruption. In doing so we also attempt to address broader questions relating to magma ascent and the changing crystallinity and rheology that will ultimately control the eruption of hydrous mafic magmas in subduction zones. The Cerro la Pilita scoria cone and lava flow Throughout the last 3 Myr, and particularly in the Quaternary, eruptions of basalt (SiO 2 552%) have been infrequent in the western Mexican volcanic belt, and of the varieties that do occur (Luhr & Carmichael 1981, sample 22E), those that contain hornblende, and thus display mineralogical evidence of the host magma being hydrous, are very rare. The trachybasalt of a small Quaternary cone (Cerro la Pilita) in the southern part of the Michoacan---Guanajuato Volcanic Field, located in the western portion of the Mexican Volcanic Belt, is the one example found to date. It consists of a scoria cone and lava flow pre-dating the historical ( ) basaltic andesites of Volcan Jorullo, Michoacan (Luhr & Carmichael 1985). Scoria of ne-normative trachybasalt ( Jor 46) are found high on the cinder cone and are conspicuous for their phenocryst assemblage of hornblende without reaction rims, apatite, olivine and augite, with scattered plagioclase microphenocrysts. On the lower slopes of the Cerro la Pilita cone are small blocks of lava (MAS 2003a) that contain phenocrysts of biotite in addition to the phenocryst phases listed above (but with less augite), and hornblende and plagioclase as unambiguous phenocrysts. The lava emitted from the cone is more siliceous than the scoria, and hornblende, without biotite, occurs therein as residual cores to extensive oxide rims. Trachybasaltic magmas (LeMaitre et al., 1989) are commonly associated with ocean island basalts and intraplate magmatism, but where they do occur in volcanic arcs, hornblende-bearing trachybasalts are comparatively common, accounting for most of the occurrences of hornblende-bearing basalts as a whole. They also tend to be associated with more K-rich shoshonitic magmas, as in the Atenguillo graben of western Mexico (Wallace & Carmichael, 1992), Papua New Guinea (Ruxton, 1966; Arculus et al., 1983), and the Hellenic Arc (Wyers & Barton, 1986; Barton & Wyers, 1991; Pe-Piper & Piper, 1992). The trachybasalt lava from Jorullo is distinctive because of its local association with basaltic andesite and its very high P 2 O 5 content (Table 1), a feature it shares with minettes (lavas with phenocrysts of biotite, apatite and olivine) from western Mexico; however, it is less rich in MgO and Ni than the absarokite lavas associated with minettes (Wallace & Carmichael, 1992; Carmichael et al., 1996). This paper represents the first attempt to explore the conditions of water-saturated phase equilibria for an erupted basaltic magma that clearly contains hornblende as part of its stable mineral assemblage. Although such a composition may not readily be considered to be representative of an `average' subductionrelated mafic magma, it is necessary to elucidate those conditions under which eruption of hornblende-bearing basalt is possible in order to understand why such magmas are largely absent at the Earth's surface. EXPERIMENTAL PROCEDURES Synthesis experiments were performed using finely powdered samples of Cerro la Pilita basalt [sample Jor 46 of Luhr & Carmichael (1985)]. Experiments were conducted between 40 and 300 MPa using an internally heated pressure vessel (IHPV) with a vertical rapid quench design (Holloway et al., 1992) with modifications as described by Moore & Carmichael (1998). Experiments were designed to explore all the possible variables that could control phase equilibria and were thus conducted in sealed Ag 70 Pd 30 capsules both under water-saturated conditions and with the fugacity of water reduced by half in the vapour phase. For concordance with the natural assemblage, 486

3 BARCLAY AND CARMICHAEL ERUPTIBILITY OF A HORNBLENDE BASALT Table 1: Summary of experimental conditions and phases for water-saturated experiments Sample no. P (MPa) T ( C) f O2 (log10) DNNO Duration (h) K D ol Phases Jor ol, S Jor ol, S Jor ol, S, Ap Jor ol, cpx, S, Ap Jor ol, cpx, S, Ap Jor ol, cpx, amph, bi, S, Ap Jor y ol, cpx, amph, bi, S, Ap Jor amph, bi, S, Ap Jor amph, bi, S, Ap Jor z 48 ol, bi, S, Ap Jor ol, cpx, amph, bi, S, Ap Jor z 45 ol, cpx, S, Ap Jor ol, cpx, amph, bi, S, Ap Jor ol, cpx, Ap, S Jor y ol, cpx, amph, S, Ap Jor z 48 ol, cpx, S, Ap Jor y ol, cpx, amph, S, Ap Jor x ol, cpx, amph, S, Ap Jor cpx, amph, S, Ap Jor ol, cpx, S, Ap Jor ol, cpx, plag, S, Ap Jor x ol, cpx, plag, S, Ap Jor { ol, cpx, plag, amph, S, Ap Jor x ol, cpx, S, Ap Jor y ol, cpx, amph, bi, plag, S, Ap Jor ol, cpx, amph, plag, S, Ap *No sensor used or sensor capsule lost water during experiment. Oxygen fugacity estimated using best-fit equation for IHPV and 0.1% H 2 mixture of Moore et al. (1995). The 95% confidence interval is equivalent to 1 log unit f O2. ylarge temperature gradient between two thermocouples monitoring sample T: Jor and 1035 C; Jor and 922 C. zexperiment has low f O2 as performed after IHPV vessel flooded with 0.3 wt % H 2 mix. xinitial experimental conditions: 138 MPa, 997 C (Jor46.32); 128 MPa, 1100 C (Jor46.29); 69 MPa, 1135 C (Jor46.28). {Run used homogeneous glass from 1 atm experiments (Table 2) as starting material. Gas replenished halfway through run to ensure stability of f O2. K D ol was calculated from exchange reaction with FeO in melt calculated using the known f O2 and expression of Kress & Carmichael (1991). Abbreviations for phases: ol, olivine; cpx, augite; amph, amphibole; bi, biotite; plag, plagioclase; S, spinel or ilmenite; Ap, apatite. experiments were largely conducted at nickel---nickel oxide (NNO) 2 but a few runs were also performed at oxygen fugacities ( f O2 ) equivalent to NNO 05 to investigate the effect of f O2 on the composition of the ferromagnesian phases, particularly the stability of hornblende. Individual run conditions and results from these three sets of experiments are summarized in Tables Additionally, four atmospheric pressure experiments were performed in a Deltech VT-31-DT gas-mixing furnace by Mo and Carmichael (Table 4, previously unpublished results) at f O2 equivalent to the NNO buffer. The sample assembly for a run in the IHPV consisted of a melt capsule and an oxygen sensor capsule connected together and hung from Pt wire. Up to 50 mg of powder was sealed into a capsule of Ag 70 Pd 30 with sufficient distilled water to achieve H 2 O saturation (water-saturated experiments) or with sufficient crystalline oxalic acid dihydrate to achieve 05H 2 O and 05CO 2 in the vapour phase during the experiments (H 2 O---CO 2 saturated experiments). The Ag 70 Pd 30 capsules were welded at both ends and samples were carefully weighed before and after each experiment and then weighed again following puncturing and 487

4 488 Table 2: Compositions of experimental samples from H 2 O---CO 2 experiments Sample no. P (MPa) T ( C) f(h 2 O) (MPa) Phase n Composition SiO 2 TiO 2 Al 2 O 3 FeO T MnO MgO CaO Na 2 O K 2 O Total Jor gl (4) 1.43(13) 19.2(1.0) 5.65(13) (0.7) 5.8(1.3) 6.19(55) 3.70(31) ol 8 Fo (4) (7) 0.28(4) 40.6(8) 0.22(4) A, S 1 ulv 28 mt Jor ol 6 Fo (12) (1.3) 0.34(7) 37.1(1.1) 0.27(9) cpx y plag 6 An 38 Ab 52 Or (2.4) (2.1) 1.1(1) (2.5) 6.1(9) 1.2(6) A, S 1 ulv 54 mt Jor ol 9 Fo (9) (4.5) (4.1) 0.20(2) cpx 12 Wo 46 En 43 Fs (2.5) 1.3(6) 4.0(1.5) 6.6(1.4) (1.2) 22.1(3) 0.59(4) plag 6 An 38 An 55 Or (5) (4) 1.05(6) 0.06(2) 0.08(1) 7.6(4) 6.5(6) 1.3(9) A, S ilm n, number of analyses; , element not analysed or below limit of detection on electron microprobe for particular phenocryst phase. *One standard deviation of replicate analyses in terms of least units cited. yphase not analysed. JOURNAL OF PETROLOGY VOLUME 45 NUMBER 3 MARCH 2004

5 Table 3: Experimental compositions from water-saturated experiments on Jor46 performed at oxygen fugacities equivalent to NNO Sample no. P (MPa) T ( C) Phase sn n Composition SiO 2 TiO 2 Al 2 O 3 FeO T MnO MgO CaO Na 2 O K 2 O NiO/SrOy Total Jor gl (20) 1.31(31) 15.05(19) 6.4(4) 0.08(7) 3.90(3) 6.53(14) 4.72(16) 3.16(4) ol 7 Fo 823 (6) 39.46(30) (6) (5) 0.24(3) 43.7(5) 0.23(3) (6) cpx 7 Wo 462 (5)En 455 (6)Fs 85 (1.8) 52.4(1.1) (28) 5.4(8) (3) 22.9(8) 0.35(18) Fz Jor gl ol (1.2) (4.7) 0.27(10) 41.3(4.3) 0.24(4) cpx amph (3) 2.6(3) 11.1(4) 11.4(9) 0.15(5) 14.4(1.0) 11.0(4) 2.81(11) 0.95(8) 0.19(5) F Jor gl ol 8 Fo 807 (4) 39.01(26) (4) 0.26(4) 41.9(3) 0.29(2) cpx Wo 475 (3)En 423 (2)Fs 102 (4) 50.3(5) 1.29(18) 3.4(4) 6.38(23) 0.11(4) 14.81(10) 23.11(20) 0.41(4) F Jor gl 53.5(3) 0.71(6) 16.54(19) 5.22(20) 0.09(3) 2.00(3) 4.19(7) 5.86(11) 3.46(8) ol cores Fo 788 (2.4) 39.1(4) (2.1) 0.32(4) 40.7(1.8) 0.21(4) cpx amph bi F Jor gl ol Fo 81 (1) 39.25(29) (1.2) 0.26(6) 42.9(1.1) 0.20(5) 0.3(1) cpx Wo 470 (0.5)En 43 (3)Fs 10 (2) 51.0(1.7) 1.1(5) 2.9(1.4) 6.48(1.2) 0.12(2) 15.3(1.2) 22.86(18) 0.46(7) F Jor gl ol 7 Fo 829 (5) 39.00(25) 16.0(5) 0.27(6) 43.51(24) 0.22(2) cpx 8 Wo 46 (1)En 43 (1)Fs 11 (1) 49.8(1.1) 1.13(28) 3.3(8) 7.0(6) 0.15(6) 14.9(7) 22.1(3) 0.55(10) F BARCLAY AND CARMICHAEL ERUPTIBILITY OF A HORNBLENDE BASALT

6 490 Table 3: continued Sample no. P (MPa) T ( C) Phase sn n Composition SiO 2 TiO 2 Al 2 O 3 FeO T MnO MgO CaO Na 2 O K 2 O NiO/SrOy Total Jor gl ol cpx Wo 46 (1)En 45 (2)Fs 9 (1) 51.7(8) 0.69(18) 2.2(6) 5.6(6) 0.10(3) 16.0(7) 22.5(5) 0.45(6) F Jor gl ol 10 Fo 87 (1) 39.89(28) (1.0) 0.24(5) 46.2(9) (4) cpx 5 Wo 46 (2)En 44 (1)Fs 10 (2) 49.9(1.5) 1.02(17) 3.3(9) 6.3(6) 0.10(3) 15.2(5) 22.60(9) 0.46(4) F Jor gl ol 6 Fo 830 (0.5) 39.64(25) (4) 0.22(4) 43.94(29) 0.24(9) (9) cpx Wo 46 (1)En 45 (1)Fs 9 (1) 51.8(1.1) 0.73(29) 2.3(7) 5.6(6) 0.13(2) 15.9(7) 22.9(5) 0.42(6) F Mineral abbreviations as in Table 1, except F (Fe---Ti oxide). n, number of analyses. Where multiple analyses were made the number in parentheses is 1s of all analyses cited in least units of average , element not analysed or below limits of detection for operating conditions of electron microprobe. *Mineral formula calculated using mol %. ynio cited for ol, cpx, F and SrO cited for amph, and BaO for analysis of biotite in Jor zfor iron---titanium oxides Na 2 O and K 2 O columns represent oxide wt % of Cr 2 O 3 and V 2 O 3, respectively. JOURNAL OF PETROLOGY VOLUME 45 NUMBER 3 MARCH 2004

7 BARCLAY AND CARMICHAEL ERUPTIBILITY OF A HORNBLENDE BASALT Table 4: Run conditions and phases for 10 5 Pa NNO phase equilibria experiments Sample no. T ( C) Phase* Mode (%) Composition SiO 2 TiO 2 Al 2 O 3 FeO MgO CaO Na 2 O K 2 O P 2 O 5 Cr 2 O 3 Total Jor46.1atm gl(6) 49.28(52)y 1.25(16) 13.78(21) 7.94(30) 8.37(11) 7.77(12) 4.44(15) 2.64(7) 1.24(13) Jor46.1atm gl(7) (38) 1.47(37) 14.30(12) 7.51(46) 7.48(15) 8.12(32) 4.58(20) 2.77(9) 1.21(25) ol(9) 3 Fo 881 (5) 40.14(19) 11.5(5) 47.4(4) 0.17(2) Jor46.1atm gl(5) (11) 1.36(20) 14.53(15) 7.19(37) 6.45(12) 8.03(20) 4.60(15) 2.83(16) 1.04(8) ol(8) 4 Fo 863 (3) 39.77(41) 13.01(33) 45.9(4) 0.40(5) sp(2) Jor46.1atm gl(3) (13) 1.30(46) 15.02(15) 6.58(27) 5.84(13) 8.52(37) 4.33(12) 2.74(9) 1.05(3) ol(7) 6 Fo 858 (2) 39.84(23) 13.43(22) 45.7(5) 0.44(14) sp(2) Jor46.1atm gl(3) (26) 1.47(10) 14.83(22) 7.68(30) 5.44(9) 8.56(7) 4.82(14) 2.83(19) 1.40(8) ol(6) 7 Fo 842 (2) 39.59(23) 14.84(16) 44.47(34) 0.43(4) spz 1 *Number in parentheses denotes number of analyses. yone standard deviation of replicate analyses in terms of least units cited. zfe---ti oxides in 1atm.9 were too small to analyse. drying ( 100 C) for 5 min to confirm the presence of a fluid phase at the end of the run. This alloy has been consistently shown to minimize iron loss in H 2 O-rich experiments (Sisson & Grove, 1993; Grove et al., 1997; Blatter & Carmichael, 1998; Moore & Carmichael, 1998) and our analysis of capsule material after the runs confirms that iron loss to the capsule was below the limit of detection (500 ppm). Pressure was measured with a calibrated Heise gauge accurate to 5 bars and temperature was monitored using three (one opposite the melt, one at the bottom of the melt capsule, one opposite the base of the sensor capsule) `K' type thermocouples calibrated to 5 C of the melting point of gold. The temperature gradient was generally less than 10 C across the melt capsule; if the gradient was greater, then this is noted in Table 1. The f O2 was buffered at approximately NNO 2 by using 01 vol. % H 2 in the argon pressurizing medium, and monitored using a H 2 O-saturated, compressed mol % Ni---Pd pellet, isolated from the Ag 70 Pd 30 capsule within an alumina sleeve. The f O2 values were calculated based on the mole fraction of Ni in the sensor capsule corrected for any differences in melt and sensor temperature (usually T melt 30 C) using the calibration of Pownceby & O'Neill (1994). The estimated error on reported f O2 values as the result of uncertainty in temperature, inhomogeneity of Ni---Pd powder, and analytical error is 02 log 10 units f O2. ANALYTICAL TECHNIQUES Experimental glasses and minerals were analysed on the Cameca SX-50 microprobe at UC Berkeley. All mineral phases were analysed using an accelerating voltage of 15 kv, and a beam current of 20 na. A focused beam was used for olivine, augite, oxides and apatite, and a slightly defocused beam (5 mm) was used for biotite, hornblende and plagioclase, to minimize alkali loss. Standards used were oxides of Mg, Si, Al, Ti, Mn, V, Cr, synthetic fayalite (Fe), diopside (Ca), nepheline (Na), orthoclase (K), strontium titanate (Sr), fluor-phlogopite (F) and chlor-apatite (Cl and P). MAN corrections (Donovan & Tingle, 1996) and 10 s counting times were used on major elements (Si, Al, Mg, Fe, Ca, Na, K). Longer count times and off-peak corrections were applied to less abundant elements (Ti, Mn, Sr, P, Cl, F) as appropriate. A natural kaersutite sample, analysed by wet chemistry (Brown & Carmichael, 1969), was used as an internal secondary standard throughout each analytical session. Tabulated analyses are the result of multiple analyses of rims and cores of several different crystals unless otherwise stated. The Ni---Pd alloy compositions of the oxygen sensors were analysed using a focused beam and a sample current of 40 na. The reported alloy compositions are the average of individual grain analyses, with 1s 53 mol % Ni. 491

8 JOURNAL OF PETROLOGY VOLUME 45 NUMBER 3 MARCH 2004 Hydrous glasses were analysed using a 5 na, 20 mm beam to minimize sodium loss. A less defocused beam and a slightly stronger beam current using an on-line Volatile Element Correction procedure (where Na, K and Si counts were monitored every 2 s) was also used, and it was shown that although the precision of these measurements was greater, the former method gave preferred results. As the result of volatile element loss, we estimate the error in Na is 10%. Using such a broad beam presented some difficulties in analysing melt pools in crystal-rich experiments. However, with the large sample sizes, this was possible in all but a few of the 50 MPa experiments. In some cases, only single analyses are presented as the number of melt pools 420 mm was limited. The analytical totals of the hydrous glasses are systematically low (Table 5). Modal calculations Mineral modes were calculated using a least-squares model (XTALFRAC, Stormer & Nicholls, 1981), assuming that the initial bulk composition of Jor 46 remained constant, which therefore represented the sum of the crystalline phases and the quenched glass. Although more practicable and accurate than point counting, this method is affected by the uncertainty in the analyses of the experimental glass and phenocryst phases. This is evident in the more crystalline experimental samples where the largest array of phenocryst phases are present, most of which have some degree of compositional zoning. To illustrate this, we compared least-squares calculations where we used: (1) the average crystal compositions; (2) average composition 1s for each of the experimental runs. This produced a maximum variation of 10% in the total crystallinity (much less for samples with 520 wt % crystals), although the relative uncertainty for individual phenocryst modes varied between 0 and 25%. We regard our estimates of crystallinity to have a maximum uncertainty of 10% but we found that the relative error associated with the abundance of individual phases can be greater, especially for the less abundant phenocryst phases. The calculated modes are given in Table 5. Mineral nomenclature The synthesized and natural calcic amphibole has nearly fully occupied A-sites and can be classified as pargasite (Leake et al., 1997). However, in keeping with earlier studies of amphibole stability in basaltic magmas (Helz, 1973; Sisson & Grove, 1993), and lacking Fe 3 determinations, this amphibole is more simply described as hornblende. Following Luhr & Carmichael (1985), we refer to the high Ca-pyroxene synthesized in the experiments as augite. Although a range of iron---titanium oxides (chrome spinel to titanomagnetite) has been synthesized, depending on the P, T and f O2, we refer to the accessory oxide phase as spinel. Using the mica classification of Deer et al. (1997), whereby molar Mg:Fe total 52:1, the synthesized mica is referred to as biotite. RESULTS Attainment of equilibrium The majority of experiments in this study are direct synthesis runs in which crystals have grown from melt. Where sufficient melt pools are available to make multiple analyses, the experimental glasses are homogeneous in composition (Table 5), with synthesized crystals generally being euhedral and equant or tabular, and in the less crystalline experiments they tend to reach sizes of 4100 mm (see Fig. 5, below). In general, we note that runs with515 wt % crystals tend to have very homogeneous crystal populations, and our calculated mineral---liquid partition coefficients (Table 1) for olivine and glass are similar to those reported by Sisson & Grove (1993); however, some zonation is inevitable in more densely crystalline experiments, and an increase in both inter- and intra-grain compositional variation is observed, as also noted by Sisson & Grove (1993) and Moore & Carmichael (1998). To test equilibrium in these cases, we ran three experiments where the T or P was initially held at higher values, followed by 48 h at the recorded run conditions (Table 1; Jor46.32, Jor46.29 and Jor46.28). We noted no difference between either the phenocryst assemblage or their compositions within these experiments (Tables ) and the direct synthesis experiments, although some olivine cores within the 50 MPa experiments are clearly relict from the initial higher T conditions. However, to test this further, we also conducted one experiment at 48 MPa and 1000 C ( Jor46.46) using a homogeneous glass from the 10 5 Pa experiments. Here, although hornblende was stable, no plagioclase was found to be present, conflicting slightly with the findings for Jor Both these experiments are very close to the stability limit of both plagioclase and hornblende, reflecting the difficulty of plagioclase nucleation when using glassy starting materials (Weber & Pichavant, 1986). Phase compositions are otherwise similar and this discrepancy was observed only in these crystal-rich experiments. A few relict grains were detected in the experiments, other than the presence of occasional chromite cores (1 vol. %) within some spinels, which may have originated from inclusions within the natural olivines. Although equilibrium cannot be unequivocally 492

9 BARCLAY AND CARMICHAEL ERUPTIBILITY OF A HORNBLENDE BASALT Table 5: Electron microprobe analyses and calculated normative compositions of experimental glasses Sample no.: Jor46 Jor46.2 Jor46.11 Jor46.20 Jor46.10 Jor46.4 Jor46.31 Jor46.21 Jor46.3 Jor46.19 Jor46.23 P (MPa): T ( C): Wt % xls ol cpx y y y hbl bi S Ap n SiO (18) (42) 50.08(73) 52.01(49) 53.1(9) 57.08(20) (59) 53.3(1.3) TiO (18) 1.71(17) 1.51(22) 0.94(22) 0.49(19) 0.42(10) (10) 1.5(3) 0.55 Al 2 O (30) 14.22(9) 15.07(6) 16.01(21) 15.61(20) 15.96(14) (23) 16.81(27) Fe 2 O 3 z FeO MnO (4) 0.10(8) 0.09(8) 0.13(9) 0.09(5) (4) 0.13(9) 0.05 MgO (8) 4.80(14) 4.46(12) 2.90(5) 2.70(10) 1.45(9) (17) 3.34(12) 0.92 CaO ) 7.68(20) 6.78(22) 4.80(10) 3.80(9) 2.70(6) (31) 6.13(29) 2.03 Na 2 O (3) 4.52(20) 4.70(14) 5.51(10) ) 5.65(31) (23) 5.4(4) 6.62 K 2 O (5) 2.92(8) 3.10(15) 3.20(5) 3.35(12) 3.98(6) (13) 3.4(11) 4.50 H 2 Ox Total Qz{ Co Or Ab An Lc Ne Di Hy Ac Ol Mt Ilm Hm *Wet chemical analysis by I. S. E. Carmichael & J. Hampel from Luhr & Carmichael (1985). ycpx used in calculation as reactant. zratio of FeO to Fe 2 O 3 calculated using method of Kress & Carmichael (1991) for measured f O2 and glass composition. xh 2 O solubility calculated for each glass composition at relevant pressure and temperature using the model of Moore et al. (1998). {Norm calculated using Magma v2.3 authored by Ken Wohletz. Wt % xls, amount of crystals in experimental charge calculated by mass balance using glass and phenocryst compositions using least-squares model of Stormer & Nicholls (1981). Abbreviations are as for Table 1. n, number of analyses; where melt pool is restricted `best' analysis is quoted rather than average composition (see analytical techniques section for discussion). Values in parentheses represent 1s variation of analysed glasses, shown in terms of least units cited. demonstrated in all experimental samples, we nevertheless believe that a close approach to equilibrium has been attained, sufficient to interpret several aspects of the evolution of the Cerro la Pilita magma. Water-saturated phase equilibria The phase stability fields for the trachybasalt have been determined for P H2 O up to 300 MPa and are plotted in Fig. 1 for experiments equilibrated at 493

10 Table 6: Experimental and natural olivine compositions 494 Sample: Jor46 1 Jor46 1 MAS 2003a Jor46.6 Jor46.11 Jor46.20 Jor46.2 Jor46.5 Jor46.10 Jor46.8 Jor46.9 Jor46.27 Jor46.24 Jor46.31 Jor46.32 P (MPa): rim core T ( C): n: SiO (3) 39.74(21) 39.4(7) 39.8(4) 40.1(3) 39.6(4) 39.2(3) 39.2(4) 40.1(2) 39.00(25) 39.34(14) 39.23(23) FeO (6) 13.54(25) 12.0(6) 10.6(1.2) 12.11(30) 14.0(1.7) 16.4(1.2) 15.8(8) 12.2(3) 16.0(5) 14.7(1.1) 14.9(4) MnO (4) 0.21(2) 0.21(2) 0.18(5) 0.24(2) 0.23(13) 0.27(7) 0.27(7) 0.19(4) 0.27(6) 0.30(3) 0.27(2) MgO (6) 45.61(23) 48.32(24) 48.8(6) 46.7(6) 45.1(1.4) 43.6(1.0) 43.2(6) 46.8(3) 43.51(24) 44.7(8) 44.85(17) NiO (6) (5) (26) 0.25(6) CaO (2) 0.25(5) 0.22(6) 0.20(4) 0.18(3) 0.20(5) 0.17(2) 0.19(1) 0.24(2) 0.22(2) 0.15(2) 0.20(1) Total Mol % Fo (7) 85.7 (3) 87.8(5) 89.2(1.2) 87.3(4) 85.1(1.9) 82.5(1.4) 83.0(9) 87.3(2) 82.9(5) 84.4(1.2) 84.3(4) Fa (7) 14.3(3) 12.2(5) 10.8(1.2) 12.7(4) 14.9(1.9) 17.5(1.4) 17.0(9) 12.7(2) 17.1(5) 15.6(1.2) 15.7(4) Sample: Jor46.26 Jor46.1 Jor Jor46.19 Jor46.29 Jor46.23 Jor46.28 Jor46.22 Jor46.30 P (MPa): T ( C): n: SiO (17) 39.89(28) 39.20(32) 40.0(4) 38.8(3) 39.73(32) 38.1(5) 39.86(31) 39.82(12) FeO 14.95(39) 12.8(1.0) 14.0(8) 17.5(4) 18.8(2.1) 13.5(1.1) 21.3(3.4) 12.5(9) 12.9(7) MnO 0.25(5) 0.24(5) 0.31(5) 0.25(3) 0.27(9) 0.28(9) 0.33(10) 0.21(6) 0.24(6) MgO 44.4(6) 46.2(9) 45.3(4) 46.3(0.6) 41.3(1.6) 45.3(1.0) 39.3(3.3) 46.3(8) 46.4(6) NiO 0.26(8) (5) (9) CaO 0.21(5) 0.22(4) 0.20(1) 0.20(3) 0.19(4) 0.22(4) 0.25(6) 0.18(4) 0.18(2) Total (2.4) Fo 84.1(5) 86.6(1.1) 85.2(8) 85.9(5) 79.6(2.4) 85.6(1.2) 76.5(4.2) 86.8(1.0) 86.5(7) Fa 15.9(5) 13.4(1.1) 14.8(8) 14.1(5) 20.4(2.4) 14.4(1.2) 23.5(4.2) 13.2(1.0) 13.5(7) JOURNAL OF PETROLOGY VOLUME 45 NUMBER 3 MARCH Representative analyses of olivine in Jor46 scoria. 2 Analysis cited with 1s from multiple analyses (1s in term of least units cited). n, number of analyses; FeO, total iron expressed as FeO.

11 Table 7: Natural and experimental hornblende compositions 495 Sample: Jor46 MAS 2003a Jor46.4 Jor46.10 Jor46.8 Jor46.7 Jor46.9 Jor46.31 Jor46.32 Jor46.26 Jor46.21 Jor46.3 Jor46.23 Jor Jor46.22 Jor46.30 Jor46.41y Jor46.44 P (MPa): T ( C): n: SiO (8) 41.8(4) 42.5(6) 41.9(6) 41.5(3) 41.7(8) 41.2(6) 40.5(18) 41.2(4) 41.4(4) 41.7(3) 41.2(5) (5) (3) TiO (4) 2.7(2) 2.2(0.2) 2.6(4) 2.8(6) 2.9(6) 3.1(5) 3.7(3) 3.2(3) 3.3(6) 3.2(12) 4.1(4) 3.95(26) 3.7(7) (7) Al 2 O (3) 11.0(4) 10.5(3) 11.1(5) 10.9(4) 11.3(6) 11.2(3) 11.7(3) 11.3(3) 10.9(2) 11.1(2) 11.4(4) 11.5(4) 11.2(4) (4) FeO (9) 9.1(3) 9.7(3) 10.4(9) 9.8(4) 9.5(3) 9.6(2) 9.0(3) 9.1(5) 9.4(5) 9.4(4) 9.1(4) 8.7(2) 9.4(4) (9) MnO (4) 0.07(2) 0.12(3) 0.13(3) 0.13(4) 0.12(5) 0.10(3) (4) 0.12(4) 0.12(3) 0.12(3) 0.12(3) 0.09(1) (5) MgO (9) 16.0(2) 15.9(5) 14.8(9) 15.4(2) 15.5(5) 15.5(4) 15.2(2) 15.5(3) 15.9(4) 15.4(3) 15.3(6) 15.6(5) 15.37(27) (1.0) CaO (2) 11.7(1) 11.4(4) 11.5(1) 11.5(1) 11.5(3) 11.77(10) 12.1(4) 11.6(2) 11.77(12) 11.6(2) 11.37(25) 11.56(21) 11.68(16) (4) Na 2 O (8) 2.64(3) 2.75(5) 2.8(1) 2.7(1) 2.78(13) 2.68(5) 2.55(8) 2.71(5) 2.66(6) 2.75(13) 2.84(11) 2.84(8) 2.68(11) (11) K 2 O (9) 1.03(4) 1.00(1) 0.8(2) 1.00(1) 0.99(15) 1.12(11) 1.09(5) 1.06(7) 1.12(12) 1.05(10) 0.92(13) 0.91(5) 1.08(6) (8) Total z Si 4 y Ti Al Fe Fe Mn Mg Ca Na K Total *Analysis taken from Luhr & Carmichael (1985). yhornblende formula calculated using ratio of Fe 2 O 3 /FeO in melt calculated from microprobe analysis of glass using Kress & Carmichael (1991). Where no glass is available, formula is not calculated. Jor46.41 and Jor46.44 are hornblendes from experiments run at oxygen fugacities equal to NNO 0.5. zincludes 0.03 Cr 2 O 3, 0.09 BaO, 0.29 F. BARCLAY AND CARMICHAEL ERUPTIBILITY OF A HORNBLENDE BASALT

12 JOURNAL OF PETROLOGY VOLUME 45 NUMBER 3 MARCH 2004 Table 8: Experimental and natural spinel compositions Sample: Jor46 Jor46 Jor46.6 Jor46.11 Jor46.20 Jor46.4 Jor46.2 Jor46.5 Jor46.8 Jor46.10 Jor46.7 Jor46.9 Jor46.24 Jor46.31 P (MPa): (a) (b) T ( C): SiO TiO Al 2 O Cr 2 O FeO MnO MgO NiO V 2 O Total Fe 2 O FeO Total Cr no Mg no Fe 3 no Sample: Jor46.32 Jor46.26 Jor46.1 Jor46.21 Jor46.3 Jor46.19 Jor46.23 Jor46.28 Jor46.29 Jor46.30 P (MPa): T ( C): SiO TiO Al 2 O Cr 2 O FeO MnO MgO NiO V 2 O Total Fe 2 O FeO Total Cr no Mg no Fe 3 no *Representative analyses of natural spinel compositions (Luhr & Carmichael 1985): (a) spinel included in olivine; (b) representative groundmass composition. NNO 2 (Table 1). Olivine is the liquidus phase at 10 5 Pa (Table 4), but our apparatus is unable to reach liquidus temperatures below 300 MPa, and thus the liquidus is poorly known. Olivine is followed in succession by spinel, apatite and augite at 200 MPa, and varies slightly in composition (Fo ) above the hornblende stability field, but becomes more Fa rich at lower temperatures. The temperatures at which spinel, augite and apatite crystallize are separated by only a few tens of degrees. Hornblende without biotite occurs 496

13 BARCLAY AND CARMICHAEL ERUPTIBILITY OF A HORNBLENDE BASALT Table 9: Experimental and natural clinopyroxene compositions Sample: Jor46 1 Jor46 1 MAS 2003a Jor46.20 Jor46.10 Jor46.5 Jor46.8 Jor46.9 Jor46.24 Jor46.31 Jor46.32 P (MPa): core rim T ( C): n: SiO (1.1) 51.40(1.3) 50.7(1.2) 51.6(1.9) 50.8(0.6) 49.8(1.1) 51.7(1.1) 50.8(1.0) TiO (27) 0.7(4) 0.85(19) 0.9(4) 0.86(13) 1.13(28) 0.68(18) 0.90(25) Al 2 O (9) 2.7(1.0) 2.7(0.7) 2.6(1.1) 2.9(3) 3.3(8) 2.2(7) 2.7(7) FeO* (9) 6.0(1.0) 5.8(5) 6.3(0.9) 6.7(0.9) 7.0(6) 5.4(9) 6.4(5) MnO (2) 0.08(2) 0.12(2) 0.11(4) 0.15(6) 0.15(6) 0.13(4) 0.14(3) MgO (7) 15.6(1.0) 15.8(6) 15.4(1.1) 15.4(4) 14.9(7) 16.1(7) 15.4(4) CaO (23) 22.34(22) 22.1(4) 22.00(28) 22.01(41) 22.1(3) 22.10(26) 22.24(28) Na 2 O (9) 0.47(7) 0.38(6) 0.47(8) 0.49(6) 0.55(10) 0.47(8) 0.46(7) Total Wo En Fs Sample: Jor46.26 Jor46.25 Jor46.1 Jor46.21 Jor46.3 Jor46.19 Jor46.29 Jor46.23 Jor46.28 Jor46.22 Jor46.30 P (MPa): T ( C): n: SiO (8) 51.7(8) 49.9(1.5) 50.5(1.2) 50.9(1.8) 51.4(0.8) 51.3(1.0) 50.3(1.8) 52.0(7) 51.0(1.0) 51.5(8) TiO (27) 0.69(18) 1.02(17) 0.9(3) 1.0(0.4) 0.86(25) 0.79(23) 1.1(5) 0.65(16) 0.81(19) 0.80(27) Al 2 O 3 2.4(8) 2.2(6) 3.3(9) 2.7(8) 2.6(1.2) 3.0(8) 2.4(7) 3.4(1.3) 2.0(6) 2.6(6) 2.1(7) FeO 6.5(4) 5.6(6) 6.3(6) 6.5(9) 6.2(1.6) 6.9(5) 6.4(7) 6.8(1.0) 5.5(6) 6.2(6) 6.2(8) MnO 0.09(5) 0.10(3) 0.10(3) 0.13(5) 0.11(6) 0.1(3) 0.13(2) 0.17(5) 0.09(3) 0.12(3) 0.12(4) MgO 15.6(5) 16.0(7) 15.2(5) 15.4(7) 16.0(9) 15.2(7) 15.7(6) 14.8(9) 16.0(4) 15.4(5) 15.9(5) CaO 22.3(3) 22.5(5) 22.60(9) 22.2(4) 22.3(7) 22.4(5) 22.3(4) 21.9(6) 22.44(13) 22.7(3) 22.3(4) Na 2 O 0.41(4) 0.45(6) 0.46(4) 0.52(10) 0.41(7) 0.51(8) 0.42(8) 0.56(9) 0.39(6) 0.48(6) 0.41(9) Total ) Wo En Fs Analyses taken from Luhr & Carmichael (1985). n, number of analyses; FeO, total iron as FeO. only over a very small temperature interval (Fig. 1), and is joined by biotite at higher pressures. Plagioclase (An ) (Table 10) is stable at P H2 O 575 MPa for T 4970 C in the absence of biotite (Fig. 1). Exploratory experiments at f O2 values equivalent to NNO 05 (Table 3) were performed to establish the effect of reducing f O2 on the Cerro la Pilita magma (Fig. 2). It can be seen in Fig. 2 that the upper limit of hornblende stability is 990 C, substantially less than its stability limit ( 1040 C) in experiments at NNO 2; olivine compositions are richer in the Fa component at the lower f O2. Curiously, hornblende without biotite is stable to 200 MPa at NNO 05 whereas it is stable only to 150 MPa at NNO 2 (Fig. 1). Amphibole compositions are more Fe rich (Table 8), but otherwise similar to those found in the more oxidizing experimental runs. Water-undersaturated phase equilibria Although solid---liquid experiments with a pure water fluid phase are simple to execute, it seems likely that 497

14 JOURNAL OF PETROLOGY VOLUME 45 NUMBER 3 MARCH 2004 Table 10: Natural and experimental plagioclase compositions Sample: Jor46 Jor46 Jor46.22 Jor46.23 Jor46.30 Jor46.29 P (MPa): T ( C): n: core rim SiO (55) 55.73(69) Al 2 O (24) 25.67(62) FeO (12) 1.06(11) CaO (23) 7.96(60) Na 2 O (10) 6.20(32) K 2 O (6) 0.73(14) SrO (10) BaO (9) Total Wt % An Ab Or Sr *Natural composition taken from Luhr & Carmichael (1985). n, number of analyses. magmas may have other volatile components dissolved in the liquid, and both CO 2 and S are plausible additions. We found no petrographic evidence for a mineral containing S in the trachybasalt scoria; accordingly, we diluted the water component of the vapour phase by using oxalic acid, which gives a fluid composition of equal amounts of H 2 O and CO 2 (Holloway & Wood, 1988). The results of three experiments are given in Table 2 and plotted in Fig. 3. Reducing the activity of water in the fluid phase results in the stabilization of plagioclase to higher pressures (4150 MPa) than in the water-saturated experimental runs, but the composition of the plagioclase remained about the same (Fig. 3), and not very different from the phenocryst composition (An 40 ; Table 10). Comparison with the natural assemblage A comparison of the experimentally determined phase equilibria with the petrological and textural features of the scoria can be used to constrain the conditions of storage of the magma. Hornblende forms euhedral phenocrysts and microphenocrysts in the scoria sample with oxidation rims absent or minute, whereas in the lava flow, hornblende is replaced by opacite almost completely. Experimental hornblendes exhibit some compositional variation as a function of pressure and temperature (Table 7) and those generated at P H2 O of MPa and temperatures between 1000 and 1020 C correspond most closely to those found in the phenocryst assemblage. One of the small differences between the natural and experimental results is K 2 O, which in the hornblende of the scoria is 026%, but in both the experiments and the ejected blocks (MAS 2003a), it is higher ( 11%) (Table 7). Hornblende compositions are more Fe rich in the lower f O2 experiments, suggesting that the natural assemblage crystallized under the more oxidizing conditions similar to NNO 2. The average composition of olivine at temperatures above the onset of hornblende crystallization is Fo 87 (Fig. 1), which corresponds within error (1% Fo) to the composition of the most magnesian olivine cores in the Cerro la Pilita scoria (Fo ; Luhr & Carmichael, 1985). CO 2 may have been only a minor component of the vapour phase, as the phenocrystic olivines compare well with those of water-saturated liquids (Fig. 1), rather than with those where CO 2 is equal to H 2 O in the vapour phase (Table 2). Cr 2 O 3 -rich spinel ( 40%) occurs as inclusions in the cores of the olivine phenocrysts (Table 8) but such Cr-rich compositions have not been reproduced within the P---T field of the experiments, although spinel with less Cr 2 O 3 ( %) is found at the highest P and T (Table 8). Otherwise all the experimental spinel approaches the titanomagnetite of the groundmass in composition. Little systematic compositional variation is found among the experimental augites, which have the general composition Wo 455 En 445 Fs 10 (Table 9). The augite phenocrysts of the scoria are zoned from Wo 47 En 455 Fs 75 cores to Wo 465 En 431 Fs 104 rims (Table 9, Luhr & Carmichael, 1985), whereas those from the biotite-bearing ejected blocks are more Mg rich (Wo 454 En 468 Fs 78 ). The experimental augites most closely resemble the phenocryst rim compositions, suggesting that the magma may have initially crystallized at higher pressures than those explored experimentally. Although biotite is absent in the scoria, it is present in the ejected blocks (MAS 2003a), and its composition is generally similar to those of the experimental biotites (Table 11), except that Na 2 O is higher, and K 2 O lower in the experimental phases; the Mg number of biotite is higher than in the coexisting experimental olivine, a characteristic feature of lavas with olivine and biotite phenocrysts (Carmichael et al., 1996). In our experiments, plagioclase feldspar that is compositionally similar to the high-sr plagioclase microphenocrysts ( An 40, Table 10) is not stable in the presence of hornblende until P H2 O is 560 MPa ( 3 wt % H 2 O) and the temperature is C (Fig. 1). We have interpreted the small size of the plagioclase microphenocrysts in the scoria, in contrast to the much larger and equant plagioclase phenocrysts in the ejected 498

15 BARCLAY AND CARMICHAEL ERUPTIBILITY OF A HORNBLENDE BASALT Table 11: Natural and experimental biotite compositions Sample: MAS 2003a Jor46.4 Jor46.10 Jor46.8 Jor46.7 Jor46.9 Jor46.31 Jor46.22 P (MPa): T ( C): n: SiO (4) 37.2(5) (5) 38.3(1.4) (45) TiO (23) 3.46(27) (28) 3.1(4) (5) Al 2 O (3) 14.55(24) (1.0) 14.79(24) (19) FeO (6) 8.58(20) (16) 9.1(6) (23) MnO (2) 0.06(1) (3) 0.05(3) (2) MgO (23) 20.02(23) (5) 18.1(1.5) (6) CaO (1) 0.08(3) (6) Na 2 O (1) 1.21(4) (16) 1.38(15) (2) K 2 O (8) 8.15(12) (14) 7.7(7) (13) Total BaO (14) 0.74 F (5) 0.32 Fe 2 O 3 y FeO Mg no *Includes 0.84 Cr 2 O 3. yratio of Fe 2 O 3 to FeO calculated using calculated ratio in glass. Fig. 1. Phase stability felds of trachybasalt (Tables 1 and 4) at NNO 2; &, water-saturated liquid composition. Boundary curves are continuous lines, and are labelled for the phases that crystallize on the low-temperature side. Dashed lines are water isopleths (wt %) calculated from Moore et al. (1998). Numbers by runs refer to the Fo content of the olivine (Table 6). Cross-hatched area represents the phenocryst (ol pyx ap hb spinel) equilibration P---T region, without plagioclase. The plagioclase stability region is below the crosshatched area. 499

16 JOURNAL OF PETROLOGY VOLUME 45 NUMBER 3 MARCH 2004 Fig. 2. Phase stability fields for water-saturated liquids at NNO 05 (Table 3), with numbers representing the Fo content of olivine (Table 3). Cross-hatched area is P---T region of equilibration of phenocryst assemblage (ol, ap, pyx, hb, spinel). Fig. 3. Composition of plagioclase in runs with P H2 O ˆ P CO2 (Table 2) showing the inferred upper stability boundaries for plagioclase. Lower curve is taken from Fig. 1 and Table 10. blocks (MAS 2003a), as resulting from decompressional crystallization on ascent, rather than as equilibration in the plagioclase stability field. The absence of biotite in the natural phenocryst assemblage of Jor 46 constrains the phenocryst equilibration of the magma to P H2 O 5150 MPa, and to water concentrations of 545 wt % (Fig. 1). Both hornblende and biotite reach a temperature maximum of 1040 C at pressures of 200 MPa, but our experiments cannot separate their stability limits above 170 MPa. Consequently, there is a limited range of P---T conditions (51000 C,570 MPa) at which plagioclase and hornblende, without biotite, are capable of coexisting in this water-saturated trachybasalt liquid at NNO 2. The presence of both biotite and plagioclase in a limited number of the eruptive products from 500

17 BARCLAY AND CARMICHAEL ERUPTIBILITY OF A HORNBLENDE BASALT Fig. 4. Contours of crystallinity (wt %) taken from Table 5 and Fig. 1, with numbers beside runs representing the calculated percentage. The anomalous position of the liquid with 23% crystals should be noted. Cerro la Pilita suggests that some of the magma passed through P---T conditions on or close to these phase boundaries during ascent. It is also worth noting that hornblende crystallization would be suppressed to significantly lower temperatures at lower f O2 values, reducing the P---T interval where hornblende could crystallize in the absence of plagioclase. Crystallinity and liquid compositions For each of the experimental runs that has an analysed glass composition (Table 5), the mode of that run has been calculated as described above, and the total amount of crystals (the crystallinity) 10% is given in Table 5. In Fig. 4, contours of crystallinity have been drawn using the values of Table 4, but the contours are not well constrained. However, their general form is correct, and show that with falling temperature and pressure, there is an increase in crystallinity; this same feature is found in the crystallinity curves of an andesite composition (Blatter & Carmichael, 2001). The experimental glass compositions (Table 5) are ne-normative at higher temperatures, becoming successively hy-normative with some of the most evolved liquids becoming q-normative at lower temperatures. This is a consequence of the precipitation of hornblende, which is very silica poor (Table 8) in contrast to the bulk composition of the lava (Table 5). These siliceous liquids are equivalent to high-k dacites according to the classifications of LeMaitre et al. (1989) and Gill (1981), respectively. Because the highest pressures of our water-saturated experiments were only 200 MPa, the compositions of these liquids will not be representative of liquids generated by partial melting of a crystalline trachybasalt (saturated or undersaturated with water) at the base of the crust (41 GPa), nor are the melt fractions (440%) small enough to represent the compositional trend of small to moderate degrees of melting of underplated trachybasalt. Nevertheless, it is evident that q-normative liquids can be generated by partial fusion of an ne-normative basalt. DISCUSSION The Cerro la Pilita trachybasalt: intensive variables and the ascent path from the mantle Comparison of the natural and experimental assemblage has provided several very important constraints on the depth of crystallization of the Cerro la Pilita magma. The absence of biotite in the Cerro la Pilita basaltic scoria provides a constraint on the depth of crystallization of the scoria phenocryst assemblage (Fig. 1). The phenocryst assemblage of hornblende, augite, apatite and olivine equilibrated between 1040 C and 970 C, with water contents between 45 and 25 wt % (Fig. 1). A combination of zoning in the augite phenocrysts, and Cr-rich spinel inclusions in the olivine, indicates that the trachybasalt magma was initially hotter than the phenocryst equilibration region of Fig. 1. The presence of biotite and plagioclase 501

18 JOURNAL OF PETROLOGY VOLUME 45 NUMBER 3 MARCH 2004 Fig. 5. Boundary curves of Fig. 1 extrapolated to higher pressures, based on the experimental results of Allen & Boettcher (1983) and Holloway & Burnham (1972) for hornblende, which is assumed to have the same stability as biotite. The calculated adiabat assumes 6 wt % H 2 O, and uses the data of Lange & Carmichael (1990), Ochs & Lange (1999) and Spera (2000) for the trachybasalt Jor 46 (Luhr &Carmichael, 1985). in the ejected blocks suggests that some portions of the magma equilibrated at greater depths and/or experienced partial degassing at some point prior to eruption. In Fig. 5, the temperature of the trachybasalt magma has been extrapolated from the phenocryst equilibration stage to the base of the crust (10 GPa), using a calculated adiabat (TVa/C p ) that varies from 27 /10 8 Pa for an anhydrous magma to 47 /10 8 Pa for one containing 6% water. It can be seen (Fig. 5) that olivine may have precipitated in the ascending magma at about 600 MPa ( 18 km) followed by augite at lower pressures before the magma entered the hornblende stability field. An ascending magma appears far more likely to enter the biotite hornblende field than it is to intersect the hornblende-only field, although only ejected blocks with biotite hornblende have been found, whereas both scoria and a 2 km long lava flow lack biotite. A hydrous magma missing the stability fields of biotite hornblende or hornblende would have phenocrysts of olivine, Cr-spinel and augite, and could suppress plagioclase, which is the characteristic phenocryst assemblage of the absarokites in western Mexico (Carmichael et al., 1996). The role of hornblende crystallization in prohibiting magma eruption The analysis of sample crystallinities also affords the opportunity to assess physical changes in the magma state during ascent and eruption. From Marsh's (1981) Aleutian modal data, magmas with more than % crystals are not found as lavas, presumably because their viscosity is too high to allow upward flow in the conduit. This limit reflects a change in the rheological regime, whereby magma behaves more like a brittle solid rather than a ductile mass, which occurs as a result of high crystal concentration within the magma. The experiments of LeJeune & Richet (1995) on a simple silicate liquid identified an abrupt change of rheology at crystal concentration of 40 vol. %, where clustered crystals start to strongly oppose shear deformation. In most silicate magmas the limit of crystal concentration beyond which magmas have an `infinite' viscosity is assumed to be 60 vol. %, the value for a monodisperse suspension (Pinkerton & Stevenson, 1992). Thus for crystal concentrations of between 40 and 60 vol. %, eruption of a magma body becomes highly unlikely; the imposition of a pressure gradient across the body is more likely to result in the migration and percolation of any remaining melt via brittle cracks rather than ductile migration of the entire body (Bagdassarov & Dorfman, 1998). It would therefore seem that an eruptible limit of 50% crystallinity would preclude trachybasalt magmas that had entered the plagioclase field (see Fig. 4) from erupting, and that unless the presence of CO 2 shifts the plagioclase field to higher pressures (Fig. 3), and hence to lower crystallinities, plagioclase-bearing 502

19 BARCLAY AND CARMICHAEL ERUPTIBILITY OF A HORNBLENDE BASALT Fig. 6. (a) Calculated modal amounts (wt %) of olivine, augite and hornblende (Table 5) at 200 MPa showing the increase of crystallinity with isobaric cooling, and the decrease in olivine and augite as hornblende increases in abundance. (b) Sketch overlays of scanning electron micrographs to show the abundance and morphology of crystals above and below the onset of hornblende crystallization. Scale bars represent 10 mm. lavas with phenocrysts of hornblende, augite, apatite and olivine will not occur. However, there is one anomalous value of crystallinity in Fig. 4 (23 wt %) that is inconsistent with the pattern of crystallinities, and it is close to the value of the natural assemblage ( 20 vol. %); perhaps there is indeed an anomalous region of low crystallinity, but the evidence to support this is weak. Holloway & Burnham (1972), in a study on the melting relations of a Kilauea tholeiite, first recognized the large increase in the amount of liquid as the hornblende breakdown temperature was exceeded. Looked at from the perspective of magma cooling, there will be a large increase in crystallinity at the onset of hornblende precipitation, and Holloway & Burnham illustrated this with two isobaric sections. In Fig. 6, we have plotted the crystallinity of the trachybasalt at 200 MPa (Fig. 1; Table 5) which clearly shows this marked increase in crystallinity, and also the decrease in olivine and augite accompanying the precipitation of hornblende; this is a reaction relationship that has also been recognized by Helz (1973), Grove et al. (1997) and Moore & Carmichael (1998). In Fig. 7, the increase of crystallinity accompanying the isobaric crystallization of hornblende in hydrous basaltic, basaltic andesite and andesitic liquids is plotted for a group of experiments taken from the literature. As a larger proportion of the components of basalt can be taken up in hornblende (regardless of the exact composition of the basalt and hornblende), in comparison with andesite, the increase of crystallinity is larger in basalts than in andesites (Fig. 7). It is postulated that the paucity of subduction-related trachybasalts, or basaltic magmas, at the Earth's surface may not be a consequence of their absence within the suprasubduction-zone lithosphere, but rather a result of the onset of high crystallinities with the precipitation of hornblende near the base of the crust. On leaving the upper mantle, an ascending hydrous magma will intersect a cooler crust, but rather than crystallize by decompression as it does on nearing the surface, it will be trapped and crystallize isobarically. 503